Shape optimisation of technical devices via gradient descent using convolutional neural network proxies

A system for optimizing the shape of a technical device adapted to be exposed to a fluid flowing around a contour of said device with respect to at least one fluid dynamic parameter, comprising:

• means for discretizing the shape of the technical device into a three-dimensional surface mesh;
• means for using a Geometric Convolutional Neural Network (CNN) to determine a prediction of the at least one fluid dynamic parameter based on the three-dimensional surface mesh, wherein the Geometric CNN comprises:
• a first branch for computing a pressure profile along the contour of the technical device (CP), and
• a second branch for computing a pressure and friction drag coefficient (CD);
• means for defining an optimization objective as a function of the shape of the technical device and of the determined at least one fluid dynamic parameter;
• and
• means for optimizing, based on the optimization function, the shape of the technical device to achieve an optimal shape.

A system for optimizing the shape of a technical device adapted to be exposed to a fluid flowing around a contour of said device with respect to at least one fluid dynamic parameter, comprising:

• a processor configured for:
• discretizing the shape of the technical device into a three-dimensional surface mesh;
• inputting the three-dimensional surface mesh into a Geometric Convolutional Neural Network (CNN) to determine a prediction of the at least one fluid dynamic parameter, wherein the Geometric CNN comprises:
• a first branch for computing a pressure profile along the contour of the technical device (CP), and
• a second branch for computing a pressure and friction drag coefficient (CD);
• defining an optimization objective as a function of the shape of the technical device and of the determined at least one fluid dynamic parameter;
• and
• optimizing, based on the optimization objective, the shape of the technical device to achieve an optimal shape.